中文摘要：

目的比较对数似然比法（log-odds法，含两种调整方法）、修正log—odds法、客观贝叶斯法估计阳性预测值区间的精密度和可靠度，探讨不同情况下的适用方法。方法以区间长度和覆盖概率为指标，比较阳性预测值区间估计的精密度和可靠度。使用SAS9．13编写MonteCarlo模拟抽样程序，完成客观贝叶斯法的计算。结果log—odds法的精密度和可靠度均低于客观贝叶斯法。大样本时，客观贝叶斯法和修正log—odds法的精密度和可靠度相似；小样本时，后者的精密度稍高，但可靠度远低于前者。结论大样本时建议使用修正log—odds法，小样本时建议使用客观贝叶斯法。

英文摘要：

Objective To compare the intervals estimated for positive predicative value by log-odds method （including two adjusting ways） modified log-odds method and objective Bayesian method in terms of accuracy and reliability, and to determine the most appropriate approach under different conditions. Methods The accuracy and reliability of interval estimation were measured by the interval length and coverage probability, respectively. All calculations relevant to objective Bayesian method were accomplished using Monte Carlo sampling technique in SAS 9. 13. Results In accuracy and reliability, objective Bayesian method was better than log-odd method, but was similar to modified log-odds method in the case of large samples. When the sample was not large enough, modified log-odds method was slightly better in accuracy, but was remarkably worse in reliability than objective Bayesian method. Conclusion modified logodds method is suggested in the case of large samples, while objective Bayesian method is preferred in the case of small samples.